FR3058230A1 - SPECTROMETRY DEVICE - Google Patents

Abstract

The invention relates to a device for spectrometry of photon radiation, the device (20) comprising: a detector (15) configured to receive the radiation and to deliver on an output (24) an electrical signal depending on the radiation (X) received, characterized in that the device (20) further comprises: • a reference base (42) parameterizable by means of a first parameter, • a comparator (38) configured to establish a comparison between the electrical signal and the reference signal, the comparator (38) delivering a signal (E (t)) representative of the energy of each photon of the radiation and a quality factor (B (t)) of the comparison, • a loop of servocontrol (46, 50, 52) for adapting the first parameter of the reference base (42) so as to optimize the quality factor (B (t)).

Description

Spectrometry device

The invention is in the field of ionizing radiation imaging, for example using X or gamma radiation. It applies to medical imaging and non-destructive screening such as baggage screening. It relates to a device for processing a photon flux.

X-ray imaging systems are used in particular for screening baggage at airports or in secure locations. Such an imaging system comprises an X-ray source subjecting the baggage to a flow of X photons and a semiconductor detector receiving the X photons having passed through the baggage. Depending on the absorption of X-rays by the luggage, it is possible to analyze the contents of the luggage. The analysis of the content of baggage and the evaluation of its dangerousness are more or less complex and automated steps depending on whether it is baggage intended for hold or baggage carried by passengers in the cabin. In both cases, the objective of X-ray imaging systems is to detect dangerous substances such as explosives and flammable materials. With the evolution of security requirements, it becomes necessary to be able to determine any type of substance in order, for example, to identify the presence of several substances which, when combined, can produce explosives.

This determination can be made by means of spectrometric imaging devices making it possible to detect each individual photon of an image (for example in X-rays), and to quantify it in energy, in order to make a classification of these photons in a histogram for build an energy spectrum of the incident flux.

X-photons are detected by a detector which converts the energy deposited by the incident photons into an electrical signal. A detector includes, for example, a crystal configured to transform the energy received from a photon into charge packets, measured by a charge amplifier. Other detectors may for example include a scintillator associated with a photodiode or a photomultiplier.

The quality of spectrometry depends mainly on the ability to best distinguish individual X photons from each other. The maximum flux of photons that the detector can process by distinguishing the different photons essentially depends on the response times of the detector and on a processing chain arranged downstream of the detector.

In existing detectors, there is a significant variation in response time due in particular to drifts when the detector is started and over time. Significant drifts are also due to variations in temperature of the detector.

To overcome this problem, use is made of a stabilization step during start-up, a step which can last on the order of an hour, or more depending on the temperature. It is also necessary to provide for regular stop periods, which is very restrictive in practice.

For some detectors, a polarization signal is applied. In the case of a detector implementing a semiconductor crystal, the polarization signal is a high voltage.

It is for example known from document FR 2 925 170 A1 to have an integrator type charge amplifier at the output of the detector. We pass the output signal of the amplifier through a delay line and we perform a subtraction of the signal leaving the delay line and the signal coming directly from the amplifier. The amplitude of the subtractor output is then a function of the energy of each photon received by the detector. This energy measurement functions correctly as long as the duration of the delay line is greater than the duration of the edge of the signal leaving the amplifier in order to distinguish by means of the subtractor the rise of the signal coming directly from the amplifier and the descent of the same inverted signal and delayed by the delay line.

On the other hand, if the duration of the delay line is less than that of the outgoing edge of the amplifier, telescoping occurs in the subtractor and the measurement of the amplitude of the pulse very significantly underestimates the real amplitude of the front of the charge amplifier output signal, and therefore the real energy deposited by the X photon. As this amounts to losing in this case part of the charges generated by the detector, the signal to noise ratio of the measurement is strongly degraded.

The correct measurement of the energy of a photon therefore requires a sufficiently long delay line duration.

Furthermore, to be able to best distinguish the individual photons from each other, a priori randomly distributed over time, and correctly quantify the energy that each has deposited on the detector, the successive pulses must be as distinct as possible, therefore the shortest possible. In other words, the higher the photon flux, the shorter the duration of the delay line in order to allow the distinction of each of the photons.

These two constraints: measurement of the energy of each photon and distinction of each photon are therefore antagonistic in terms of the choice of the duration of the delay line. In practice for a large photon flux, typically more than 1 million per second, it becomes impossible to determine a delay line duration allowing both to sufficiently distinguish the individual photons (resolution of coincidences), while correctly quantifying the energy that everyone has deposited on the detector.

The invention aims to overcome all or part of the problems mentioned above by proposing a spectrometry device making it possible to stabilize the detector or at least to correct the effects of the instability of the detector.

To this end, the subject of the invention is a device for spectrometry of photon radiation, the device comprising:

• a detector configured to receive the radiation and to deliver an electrical signal according to the received radiation to an output, characterized in that the device further comprises:

• a configurable reference base by means of a first parameter, • a comparator with two inputs and two outputs, the comparator receiving on its first input the electrical signal, and on its second input a reference signal delivered by the reference base , the comparator being configured to establish a comparison between the electrical signal and the reference signal, the comparator delivering on its first output a signal representative of the energy of each photon of the radiation and on its second output a quality factor of the comparison , • a control loop allowing the first parameter of the reference base to be adapted so as to optimize the quality factor.

In a first embodiment, the device further comprises an integrator-type amplifier comprising an input and an output, the input of the amplifier being connected to the output of the detector. The reference base is then configured to generate several forms of signals, each corresponding to a front duration of a signal delivered by the amplifier in response to the reception of a photon by the detector.

The reference base is advantageously configured to associate with each form of signals a reference edge duration forming the first parameter of the reference base, the control loop slaving the reference edge duration to the edge duration of a signal delivered by the amplifier in response to the photon being received by the detector.

The signal representative of the quality of the comparison is advantageously representative of a difference between the edge duration of a signal delivered by the amplifier and the reference edge duration associated with the form of signals received by the comparator. The loop for controlling the reference edge duration to the edge duration of a signal delivered by the amplifier is configured to modify the reference edge duration in order to change the reference shape as a function of the value of the representative signal. the quality of the comparison.

The control loop of the reference edge duration to the edge duration of a signal delivered by the amplifier may include a multiplier of the representative value by a gain constant and an adder adding a result from the multiplier to the current value of the reference edge duration.

In the first embodiment, the device can further comprise:

• a subtractor with two inputs and an output, connected to a first of its inputs at the output of the amplifier, the output of the subtractor delivering the electrical signal, • a delay line connected between the output of the amplifier and a second of the two subtractor inputs ,.

The delay line can be parameterized by means of its delay time and the reference base can be parameterized by means of a second parameter formed by the delay time.

In a second embodiment, the reference base delivers a time filter parameterized by a time constant. The comparison consists in applying the filter to the electrical signal, the signal representative of the quality of the comparison being a difference of a rise time of the filtered signal with a reference rise time, the control loop tending to cancel the signal. representative of the quality of the comparison.

The detector can be polarized by means of a polarization signal and advantageously the device comprises a loop for controlling the polarization signal as a function of a difference between a first current parameter and a first nominal parameter.

The control loop of the polarization signal can comprise a multiplier of the difference between the first current parameter and the first nominal parameter by a gain constant and an adder adding a result from the multiplier to the current value of the polarization signal.

A time constant of the feedback loop of the bias signal is advantageously stronger than a time constant of the feedback loop of the first parameter of the reference base.

The invention will be better understood and other advantages will appear on reading the detailed description of an embodiment given by way of example, description illustrated by the attached drawing in which:

FIG. 1 represents an X-ray imaging system for checking baggage;

FIG. 2 schematically represents an example of a spectrometry device implementing the invention and which can be installed in the system of FIG. 1;

FIGS. 3a and 3b show examples of a bundle of curves representing the amplitude on the ordinate of an output signal from a subtractor of the device;

FIG. 4 represents an example of a beam of digitized reference curves;

FIG. 5 schematically represents another example of a spectrometry device implementing the invention.

For the sake of clarity, the same elements will have the same references in the different figures.

FIG. 1 schematically represents an X-ray imaging system 10 for checking baggage. The imaging system 10 comprises an X-ray source 11, a conveyor belt 12 capable of transporting luggage 13 in the direction represented by the arrow 14, a semiconductor detector 15 and a processing device 16 connected to the detector 15 The detector 15 represented in FIG. 1 is a linear detector in L. It comprises elements sensitive to X-ray radiation such as photodiodes or photoconductors arranged on two lines perpendicular to each other. The detector 15 can also be a two-dimensional detector and include a matrix of sensitive elements. When a sensitive element of the detector 15 is exposed to an X photon, it converts this X photon into a packet of charges, the number of which is proportional to the energy of the photon. The X-ray source 11 emits a stream of photons towards the detector 15 through the conveyor belt 12. The spectrum of the X-ray stream received in the absence of objects is determined before analyzing the objects. Thus, by comparing this spectrum with the spectrum of the X-ray flux received in the presence of an object to be analyzed, it is possible to determine the X-ray flux absorbed by the object to be analyzed.

It is understood that the invention can be implemented in any form of photon radiation detectors such as Gamma radiation, and in other systems requiring spectrometry of photon radiation, such as for example medical imaging. .

FIG. 2 schematically represents an example of a spectrometry device 20 according to the invention and installed in the system 10.

The device 20 comprises a detector 15 and a processing chain 22 installed in the processing device 16. The detector 15 may comprise one or more sensitive elements. The device 20 can either comprise a single processing chain 22 for all of the sensitive elements of the detector 15, in which case the charge packets originating from the different photosensitive elements are treated successively by the chain 22, or comprise a processing chain associated with each photosensitive element, in which case the charge packets are processed simultaneously by the different chains 22. In order not to overload the figure, we will take the example of a detector 15 comprising a single sensitive element which will be assimilated to the detector 15 which has an output 24 on which the electrical charge packets are available to be read.

The processing chain 22 can be produced in the form of an integrated circuit suitable for processing the charges from the detector. This type of integrated circuit is for example known in the Anglo-Saxon literature under the name of "ASIC" for: "Application-Specific Integrated Circuit". It is also possible to use a programmable logic circuit notably known in the Anglo-Saxon literature under the name of FPGA for "FieldProgrammable Gâte Array" or PLD for "Programmable Logic Device". One can also implement a digital signal processor known in the Anglo-Saxon literature under the name of DSP for: "Digital Signal Processor". More generally, any circuit dedicated to signal processing can be implemented to produce the processing chain 22.

The detector 15 may comprise a crystal allowing the conversion of photons into charges, such as in particular crystals of cadmium telluride (CdTe) or zinc-cadmium telluride (CdZnTe or CZT). These crystals convert the energy of photons into electrons. Alternatively, the detector 15 may include a scintillator associated with a photodiode or a photomultiplier. The scintillator converts the incident photon radiation into another longer wavelength photon radiation to which the photodiode is sensitive. It is for example possible to use rapid scintillators based on lutetium, as known lutetium oxyortho-silicate is the abbreviation "LSO" or known lutetium-yttrium oxyortho-silicate is the abbreviation "LYSO >>.

The photomultiplier can be silicon-based, for example those known by the abbreviation SiPM for its initials in the Anglo-Saxon literature: "Silicon PhotoMultiplier" or MPPC: "Multi-Pixel Photon Counter".

FIG. 2 represents a spectrometry device whose detector implements a crystal. In this variant, the processing chain 22 comprises an amplifier 26 of the integrator type comprising an input 25 and an output 27. The input 25 is connected to the output 24 of the detector 15, optionally by means of a decoupling capacitor C1. Another capacitor C is placed between input 25 and output 27. It is possible to add a resistor in parallel with capacitor C which forms a feedback ensuring the integrator function of amplifier 26.

The processing chain 22 may include a subtractor 30 associated with a delay line 34. A positive input 31 of the subtractor 30 is connected directly to the output 27 of the amplifier 26 and a negative input 32 subtractor 30 is connected to the output 27 of the amplifier 26 via a delay line 34. At the output 35 of the subtractor 30, there is a signal formed by the difference between the signal from the amplifier and the same signal delayed by the delay line. It is possible to insert a gain upstream of one of the inputs 31 and 32 of the subtractor making it possible to balance the levels of the signals received by the subtractor 30, in particular in the event that the delay line generates a weakening of the signal. The negative and positive terminals of the subtractor 30 can be reversed. The rest of the treatment will take this reversal into account.

According to the invention, the processing chain 22 comprises a comparator 38 with two inputs 39 and 40 and an output 41. The input 39 receives the output signal from the subtractor 30 and the input 40 is connected to a reference base 42 of signal shapes. The reference base 42 comprises several forms of signals expected at the output of the subtractor 30. The signal forms correspond to different response times of the crystal of the detector 15. By choosing for the signal forms contained in the reference base 42, different times response in an expected range and by distributing them with a constant step, it is possible to find the real energy level of the photon received by the detector 15 thanks to a proximity between the real response time and one of the base signals reference 42.

The comparator 38 is configured to establish a correlation between a signal received from the subtractor 30 and one of the signals from the reference base 42. Correlation is a form of comparison carried out by the comparator 38.

FIGS. 3a and 3b represent two bundles of curves representing on the ordinate the amplitude of a signal U _p (t) output from the subtractor 30 as a function of the time represented on the abscissa. In these two figures, the amplitude has been normalized to 1 and the time is expressed in nanoseconds. In FIG. 3a, the duration of the delay line is fixed at 50 ns and each of the curves corresponds to the duration of a voltage front leaving the amplifier 26. The first curve corresponds to a voltage front during 10 ns and the last curve corresponds to a tension front during 80 ns.

The voltage front is due to the detection of a photon by the detector 15. The duration of this front is a function of the pass band of the amplifier 26, as well as the kinetics of arrival of the charges at the input of the amplifier 26. This kinetics depends in particular on the characteristics of the detector 15, on its stabilization, on the temperature, on its supply voltage. FIG. 3a illustrates that if the duration of the voltage front is greater than the duration of the delay line called Tlar, the signal leaving the subtractor 30 cannot reach its maximum amplitude: 1 in FIG. 3a.

Similarly, FIG. 3b the duration of the front is fixed at 60 ns and the different curves represent different values of the duration of the delay line 34. The same observation as from FIG. 3a can be made from the FIG. 3b, By choosing the duration Tlar less than the duration of the edge, the maximum amplitude of the signal is not reached.

Nevertheless, by storing in the reference base 42 different forms of signals corresponding to different durations of the voltage front and possibly to different durations Tlar of delay line, it is possible to compare the signal coming from the subtractor 30 with the reference signals for find the best correlation and thus go back to the energy of the photon received by the detector 15. This comparison can give a good result even if the duration of the voltage front is greater than the duration of the delay line.

The reference forms can be theoretical forms, for example calculated from operating models of the detector 15 and of the various components of the chain 22. It is also possible to define the reference forms empirically using measurements made on device 20.

The reference signals stored in the reference base 42 can be digitized in steps, for example in steps of 10 ns. A sampling of the signal U _p (t) from the subtractor 30 is then carried out at the same step. The comparison then consists in carrying out, at each sampling period, a statistical correlation calculation between the successive samples of the signal from the subtractor 30 and one of the reference forms stored in the base 42.

FIG. 4 represents an example of a bundle of digitized reference curves. This bundle of curves is similar to that of FIG. 3b. Digitization was done with a 10 ns step. Scanning spans N steps, 17 steps in this example.

The correlation calculation between the signal from the subtractor U _p (t) and the reference forms can for example be akin to a linear regression calculated at each sampling period. For the reference form F (n) chosen, n representing the sampling step of the reference form and varying from 1 to N, the following two constant values are calculated:

• An average of the reference form:

n = l • A variance of the reference form:

N n = l

The mean F and the variance S _E can be calculated in the reference base 42.

To process the pulse signal U _p (t) coming from the subtractor 30, the following calculations are carried out in parallel with each value of the time t, an integer multiple of the sampling period T _ECH which is in the example represented by 10 ns:

• Average at time t of U _p :

U,

U _p (t - (N - rf) T _ECH ) n = l

Average at time t of the square of U _p :

N v} (t) = i - («- nyr _EC „ yn = l

Variance at time t of U _p :

sèy-) = uï (t) - (ügt)) ²

For each of the reference forms: Average at time t of the product FU _P :

N ^f - ^u p ^(t) = Σ ^{F (n)} - ^{Up (t nyrEcH)} n = l

Covariance at time t of F and U _r

S _FUp (t) = FU _p (t) - F .U _p (t)

We deduce for each time t, multiple of T _ECH , a correlation coefficient R (t) function of the time between U _p (t) and F (t):

Spilflt). | SFJL ₇ (t) |

F (t) = ^? FVp (0 ppt / ps ² p. (T)

The analysis of variations in R (t) as a function of time t makes it possible to detect the arrival of a front, therefore of a photon. R (t) (dimensionless) is at any time between -1 and +1. We consider that an edge is detected each time R (t) has a relative maximum (as a function of t) greater than a given threshold RS which is for example equal to 0.75. The reference curve F having the largest value of R (t) is then retained. In practice this value is close to 1.

The means, variance and covariance as well as the correlation coefficient involving the Up signal can be obtained in the comparator 38.

The following three parameters are then calculated for each time t, a multiple of T _ECH :

• A (t) represents the variation of the amplitude of U _p (t):

• B (t) represents the ordinate at the origin of the amplitude:

B (t) = ËÇ (t) - A (t) .F • E (t) represents the energy of the photon:

E (t) = A (t) + B (t)

The amplitude of the front is then equal to E (t), t being the instant of the relative maximum of R (t). In FIG. 2, the main functional blocks allowing the calculations are shown. The calculation of A (t) is carried out in a block 45 and the calculation of B (t) is carried out in a block 46. The blocks 45 and 46 use the information delivered by the comparator 38 at its output 41. A summator 48 receives the outputs of blocks 45 and 46 and calculates E (t).

An integrator 49 receives, for each photon detected, the value E (t). The integrator 49 integrates in a histogram all the values of E (t) to obtain an energy spectrum of the photons received.

These calculations make it possible to correctly quantify the amplitude of the edges, therefore the energy of the incident photons, even if the duration Tlar of the delay line 34 is less than the edge duration T _F ref used for the reference forms F.

In addition, the calculations take into account the entire Up (t) pulse, instead of simply detecting the maximum value of this signal as in the prior art, the signal / noise ratio is further improved.

It is possible to simplify the calculations, in particular in order to simplify the processing chain 22. To do this, it is possible to directly search for the relative maximums of E (t) as a function of t. Each maximum corresponds to a detected pulse, the front amplitude of which is given by the value of this maximum. This avoids having to do the calculations of R (t), and of U ² (t) and of Sy _p (t).

We can possibly further simplify the formulas by an approximate calculation assuming that B (t) = 0 a priori. The calculation is simpler but gives less good results.

The correlation making it possible to recognize the reference signal corresponding to the energy of the photon received by the detector 15 is all the better achieved since the start of the front coming from the amplifier 26 corresponds well to a zero value of the amplitude. Generally, the charge amplifier 26 does not transmit the DC component of the signal formed by the detector 15, in particular due to the presence of the decoupling capacitor C1, useful for overcoming the leakage current of the detector 15. Consequently, the zero amplitude in the curves as shown in Figures 3a and 3b is not known. It is therefore necessary to make an offset correction of the amplitude often called, baseline correction. This correction can be done for example by averaging, possibly sliding, on successive values of the signal between two pulses. It may also be necessary to compensate for the cut-off frequency of the charge amplifier. This cutoff frequency is due to capacitor C and its series resistance present in the feedback of amplifier 26.

There remains a measurement error if the duration T _F front edge retained for the reference form with which the calculations were made differs from the duration T _F sig front edge at the output of amplifier 26. This error is particularly important when Tlar <T _F ref- More precisely, if the duration Tfsig of the edge at the output of the amplifier 26 varies over time, in particular with the temperature, the measurement loses stability. Advantageously to reduce this error, the processing chain 22 is configured so that the reference base 42 generates several forms of signals, each corresponding to an edge duration Tfsig of the signal delivered by the amplifier 26 in response to the reception of a photon by the detector 15.

The duration T _FREF represents a first parameter of the reference base 42. In addition, the delay line 34 can also be configurable by means of its delay time Tlar. Advantageously, the reference base 42 is also configurable by means of a second parameter formed by its delay time Tlar. Access to the reference form retained for the correlation is then done in a matrix fashion as a function of two parameters T _FREF and Tlar.

It was noted that the value of the parameter B (t) representing the ordinate at the origin of the signal U _p (t), is representative of the adequacy between the duration T _F sig of edge at the output of the amplifier 26 and duration T _F headline ref of the reference forms. In other words, the parameter B (t) represents a quality factor of the correlation. More precisely, if the value of B (t) at the instant t of detection of a pulse is positive, the edge duration T _FREF of the reference shape is shorter than the edge duration T _FS ig of the signal originating of amplifier 26. In other words, if B (t)> 0, then T _FREF <T _FS ig- Conversely, if the value of B (t) at the instant t of detection of a pulse is negative, the edge duration T _FREF of the reference shape is longer than the edge duration T _FS ig of the signal from the amplifier 26. In other words, if B (t) <0, then T _FREF > T _F sig When an edge is detected, more precisely using the calculations proposed above, when R (t) has a relative maximum as a function of t greater than the given threshold RS, the edge duration T _FREF of the form is corrected of reference by means of an adder 50. More precisely, to the current value of T _FREF , one adds the algebraic value of B (t) affected by a coefficient k _FAA forming a constant d e positive gain. In other words :

T FREF = T _FREF + k _FAA xB (t)

A multiplier 52 makes it possible to multiply the coefficient k _FAA by B (t). More generally, the processing chain 22 comprises a control loop from the reference _edge duration T _FREF to the edge duration T _FS ig- In the example shown, the control loop comprises the determining block 46 B (t) forming an error signal of the control loop, the multiplier 52 applying to B (t) the gain k _FAA and the adder 50 correcting the reference edge duration T _FREF with B (t) affected of his gain k _FAA .

The coefficient k _FAA is to be adjusted according to a stability / speed compromise. The speed of convergence from T _FREF to T _FS ig must be proportional to the photon flux received. However, too high a speed risks making the value of T _{FREF unstable} . At startup, T _FREF will be initialized to a typical value. Each time its value changes (at least significantly), the reference forms F (n), as well as F and Sp, are recalculated. It is possible to keep the reference forms if the value of T _FREF does not change significantly. In other words, as long as the value of T _F ref does not exceed a predetermined threshold around its current value, the reference forms remain unchanged.

It is possible to refine the modification of the current value of T _F ref, by applying a proportional, integral and / or derivative type correction. The coefficients of this type of correction are determined in order to improve the stability / speed compromise.

When the duration of the reference shape edge T _F ref has converged, it will conform to the average duration of the signal edge duration T _F sig- If the variations of T _F sig are slower than the convergence speed of T _F ref, defined by the coefficient k _F AA, then T _F ref will follow at all times the variations of T _F sig, and the quantization of energy E (t) will remain precise.

As mentioned above, the detector 15 can comprise several sensitive elements and the processing chain 22 can be common to different sensitive elements of the detector 15. However, when there are several sensitive elements (or pixels), the value of T _F ref can be specific to each sensitive element, or common to several sensitive elements. If the characteristics of the different sensitive elements are close enough, it is possible to pool the same value of T _F ref for the different sensitive elements. This allows faster convergence by pooling the photons detected by the different sensitive elements.

Advantageously, an authorized range of values of E (t) is defined, with a minimum above the background noise, and possibly also a maximum to avoid taking into account false pulses coming from photon stacks often called coincidences and received almost simultaneously by the detector 15.

The adaptation of the reference edge duration T _F ref allows, as long as there is a photon flux, a permanent self-adaptation of the duration of the edge of the reference shape so that it best matches to that of the front of the real signal. Optimal fidelity is thus obtained from the quantification of the amplitude of the front and therefore from the measurement of the energy of the incident photons.

The adaptation of the reference edge duration T _F ref makes it possible to stabilize the measurements made by the device when the response time of the detector 15 and of the processing chain 22 changes. It is possible to further improve the measurements by tending to stabilize the response time itself.

For its operation, a polarization signal is applied to the detector 15. In the case of a semiconductor crystal, the polarization signal is a voltage which can typically range from a few hundred volts to 2 kilovolts. More precisely, a received photon generates charges in the crystal and the bias voltage allows the transit of the charges generated towards electrodes. The bias voltage is called Uht or high voltage. In practice, the signal edge duration T _F sig changes as a function of the electric field applied to the crystal due to the presence of the high voltage Uht · More precisely, the signal edge duration T _F sig mainly depends on the electric field in the crystal. At constant bias voltage, there are variations in the electric field, especially when the device is started.

To keep the signal edge duration T _E sig as close as possible to a nominal reference duration T _FREF0 , It is possible to adjust the high voltage Uht - In other words, the processing chain 22 comprises a second loop of high voltage control Uht θη function of a difference between the reference edge duration T _FRFF and the nominal reference edge duration T _FREF0 .

Previously, we tried to act on the high voltage Uht- But only to try to accelerate the stabilization of the detector, not to keep constant the duration of the signal edge T _E sigThis second control loop of the high voltage Uht s 'adds to the first control loop reference signals described above. Not only, the amplitude of the signal is determined by correlation to a reference signal chosen as a function of the real edge duration T _E sig and in addition, one acts on the detector itself to maintain the real edge duration T _E sig as constant as possible.

Alternatively, it is possible to implement the second control loop without the first loop. More precisely, the polarization signal can be controlled on a variable representative of the real edge duration T _FS ig- The variable can be a measure of the rise time of the signal from the processing chain 22. It is also possible to set implement the reference base 42 with a single signal and use the quality factor of the correlation, in this case the parameter B (t) to control the high voltage UthThe second control loop, includes a comparator 55 of the reference edge duration T _F ref and of the nominal reference edge duration T _FREF0 . At the output of comparator 55, there is an error signal ε from the second control loop. A multiplier 57 makes it possible to multiply the error signal ε with a servo coefficient k _HT AA · The voltage Uht is corrected by means of an adder 59 adding to a current value of the voltage Uht a correction equal to the result of the multiplication. More precisely, to the current value of Uht, we add the error signal ε affected by the coefficient k _H TAA- In other words:

Uht = Uht + kHTAA (T _F re _F - Tfreto)

In practice, Uht can represent a pilot value proportional to the bias voltage actually applied to the detector. It is possible to refine the modification of the current value of Uht, by applying a correction k _H TAA of proportional, integral and / or derivative type. The coefficients of this type of correction are determined in order to improve the stability / speed compromise.

The second control loop makes it possible to maintain good quantification of energies under all circumstances, even during the detector stabilization or temperature stabilization phases. The only limitation is the voltage range Uht acceptable for the proper functioning of the detector 15.

The choice of the target nominal value T _FREF o can be made definitively as a function of the detector 15 chosen. Alternatively, to further improve the performance of the device relative to the physical characteristics of the detector 15, it is also possible to use a target value T _FREF o which depends for example on the ambient temperature, the high voltage Uht applied, or other physical parameters linked to the detector 15.

The two control loops can interfere with each other and cause instability of the energy measurements E (t).

Advantageously, a time constant of the second control loop is stronger than that of the first loop. In other words, from a target nominal value T _F refo of reference shape edge duration, and we will gradually adjust the absolute value of voltage Uht, until converging so that T _F ref - Tfrefo is zero on average over a long period of time. The modification of series of reference shapes carried out by means of the first control loop has a lower reaction time.

FIG. 5 schematically represents another example of a spectrometry device 70 according to the invention and which can be installed in the system 10. In the device 70, there is a detector 15 and the associated processing chain 22. The reference base here bears the reference 72. It is a temporal filter, for example a Gaussian filter whose time constant τ is configurable. The comparator here is a convolution operator 74 making it possible to apply the filter to the signal from the processing chain 22. In other words, the comparison carried out by the comparator consists in applying the filter to the electrical signal coming from the detector 15. The convolution is a form of comparison. Block 76 corresponds to blocks 45 and 46 described above and making it possible to calculate the quality factor B (t) and the energy E (t). The quality factor B (t) is for example a difference between the rise time of the filtered signal from the comparator 74 and a reference rise time. The energy E (t) is for example the amplitude of the filtered signal. In the first control loop, a block 78 corresponding to the multiplier 52 and to the adder 50 makes it possible to adapt the time constant τ of the reference base 72 to obtain the optimal value of the quality factor, in this case, a zero value of B (t). here also a coefficient k _F AA is to be adjusted as a function of a stability / speed compromise. As in the first example, it is possible to refine the value of the time constant τ by applying a proportional, integral and / or derivative type correction. The coefficients of this type of correction are determined in order to improve the stability / speed compromise.

As in the first example, a second control loop can act on a polarization signal from the detector 15 so as to keep the time constant τ as close as possible to a reference time constant τ ₀ . A comparator 80 corresponds to comparator 55 and a block 82 takes over the functions of multiplier 57 and adder 59.

Claims (10)

1. A device for spectrometry of photon radiation, the device (20) comprising: • a detector (15) configured to receive the radiation and to deliver an output (24) an electrical signal depending on the radiation (X) received, characterized in that the device (20) further comprises: • a reference base (42) configurable by means of a first parameter, • a comparator (38) with two inputs (39, 40) and two outputs (41), the comparator (38) receiving on its first input (39 ) the electrical signal, and on its second input (40) a reference signal delivered by the reference base (42), the comparator (38) being configured to establish a comparison between the electrical signal and the reference signal, the comparator (38) delivering on its first output a signal (E (t)) representative of the energy of each photon of the radiation and on its second output a quality factor (B (t)) of the comparison, • a loop of servo-control (46, 50, 52) allowing the first parameter of the reference base (42) to be adapted so as to optimize the quality factor (B (t)). 2. Device according to claim 1, characterized in that it further comprises: An amplifier (26) of the integrator type comprising an input (25) and an output (27), the input (25) of the amplifier (26) being connected to the output (24) of the detector (15), and in that the reference base (42) is configured to generate several forms of signals, each corresponding to an edge duration (Tfsig) of a signal delivered by the amplifier (26) in response to the reception of a photon by the detector (15). 3. Device according to claim 2, characterized in that the reference base (42) is configured to associate with each form of signals a reference edge duration (T _F ref) forming the first parameter of the reference base (42 ), the control loop (46, 50, 52) slaving the reference edge duration (T _FREF ) to the edge duration (Tfsig) of a signal delivered by the amplifier (26) in response to reception of a photon by the detector (15). 4. Device according to claim 3, characterized in that the signal (B (t)) representative of the quality of the comparison is representative of a difference between the edge duration (Tfsig) of a signal delivered by the amplifier (26) and the reference edge duration (T _F ref) associated with the form of signals received by the comparator (38) and in that the control loop for the reference edge duration (T _F ref) at the edge duration (Tfsig) of a signal delivered by the amplifier (26) is configured to modify the reference edge duration (T _F ref) in order to change the reference shape as a function of the value of the signal (B ( t)) representative of the quality of the comparison. 5. Device according to claim 4, characterized in that the control loop from the reference edge duration (T _F ref) to the edge duration (Tfsig) of a signal delivered by the amplifier (26) comprises a multiplier (52) of the representative value (B (t)) by a gain constant (kpAA) and an adder (50) adding a result from the multiplier (52) to the current value of the reference edge duration (T _F ref) · 6. Device according to one of claims 2 to 5, characterized in that it further comprises: • a subtractor (30) with two inputs (31,32) and an output (35), connected to a first of its inputs (31) at the output of the amplifier (26), the output (35) of the subtractor ( 30) delivering the electrical signal, • a delay line (34) connected between the output (27) of the amplifier (26) and a second of the two inputs (32) of the subtractor (30), and in that the line delay (34) is configurable by means of its delay time (Tlar) and in that the reference base (42) is configurable by means of a second parameter formed by the delay time (Tlar). 7. Device according to claim 1, characterized in that the reference base delivers a temporal filter parameterized by a time constant, in that the comparison consists in applying the filter to the electric signal, the signal (B (t)) representative of the quality of the comparison being a difference of a rise time of the filtered signal with a reference rise time, the control loop tending to cancel the signal (B (t)) representative of the quality of the comparison. 8. Device according to one of the preceding claims, characterized in that the detector (15) is polarized by means of a polarization signal (Uht) and in that the device comprises a control loop (55, 57, 59) of the polarization signal (Uht) as a function of a difference between a first current parameter (T _F ref) and a first nominal parameter (T _FREF0 ). 9. Device according to claim 8, characterized in that the control loop of the polarization signal (Uht) comprises a multiplier (57) of the difference between the first current parameter (T _FREF ) and the first nominal parameter (T _FREF0 ) by a gain constant (Khtaa) and an adder (59) adding a result from the multiplier (57) to the current value of the polarization signal (Uht) · 10. Device according to any one of claims 8 or 9, characterized in that a time constant of the control loop (55, 57, 59) of the polarization signal (Uht) is stronger than a constant time of the control loop (46, 50, 52) of the first parameter (T _FREF ) of the reference base (42).